Industrial Organization
Health and Pharmaceutical Economics
Applied Micro-Econometrics
International Trade and Development

Industrial Organization
Health and Pharmaceutical Economics
Applied Micro-Econometrics
International Trade and Development

Industrial Organization and Microeconomic Theory (Oral), 2000
Microeconomic and Macroeconomic Theory (Written), 1999

Pharmaceutical Demand and Welfare Implications of Innovation

Professor Steven Berry
Professor Fiona M. Scott Morton
Professor Penny Goldberg

May 2003

Ph.D., Yale University, expected May 2003
M.Phil., Yale University, December 2001
M.A., Yale University, May 2000
B.S., University of Michigan, Ann Arbor, May 1998
     High Distinction & Highest Honors, Economics and Mathematics

Economics Department Fellowship, Yale University, Summer 2002
John Perry Miller Fund Award, Yale University, September 2001
Yale University Dissertation Fellowship, Fall 2001
Yale University Graduate Student Fellowship, 1998-2002
Osterweil Prize in Economics (most outstanding and promising student), U. Michigan, May 1998
Phi Kappa Phi, admitted 1998
Phi Beta Kappa, admitted 1997
Angell Scholar (academic achievement award), U. Michigan, 1998
Honors Program Thesis Research Award, U. Michigan, Ann Arbor 1997
Class Honors (academic achievement award), U. Michigan, 1996-1998
Branstrom Prize, (academic achievement award), U. Michigan, 1996
Fulbright Scholar, 1995-1998

Head Teaching Assistant: Introductory Microeconomics (Fall 2002)
Teaching Assistant: Industrial Organization, Mathematics for Economists (Ph.D.), Introductory Microeconomics, Intermediate Microeconomics (2000-2002)
College Tutor: Economics, Statistics and Mathematics, Silliman College (2001-2002)

Research Assistant to Professor Steven Berry, (2000-2001), Yale University: Economic research and programming on hedonic pricing models
International Monetary Fund, Summer (2000), Geneva, Switzerland: Research on the costs and benefits of WTO membership for emerging markets and LDC/HIPC countries
Research Assistant to Professor David Weinstein, Summer (1998), University of Michigan Business School: Research on International Trade in the Manufacturing Industry
Visiting Researcher, Summer (1996), Institute for Advanced Study, Princeton: Research on topics of modern probability theory (random walks) and applications

"Patient Welfare Implications of Innovation and Diffusion in the U.S. Antidepressant Market," September 2002 [job market paper]

"Analyzing the Effects of Marketing Efforts by Pharmaceuticals on Patient Welfare," in progress

"Rising Drug Costs and the Strategic Use of Exclusivity Rights Post Patent Expiration: Would Industry Regulation be Justified?" in progress

"Does Openness Lead to More or Less Development? The Case of Health Deterioration" Yale University, May 2001

"An Empirical Economic Analysis of GATT/WTO Membership for Selected Countries," IMF and Yale University, November 2000

"In Search of the Effects of Tariffs on Bilateral Trade: Does the Number of Source Countries Matter?" Yale University, December 1999

"European Monetary Union: A Union of Countries or a Union of Regional Optimum Currency Areas," University of Michigan, Ann Arbor, May 1998 [undergraduate honor thesis]

English, Greek, French; elementary knowledge of Spanish and Russian

Returns to innovation constitute a major component of social welfare. To evaluate innovations and their diffusion over time, requires use of a methodology that isolates their precise effect on welfare. This is particularly important in the pharmaceutical industry where rapid innovation occurs and drug development costs are high. In this study, I formulate an empirical methodology that quantifies patient welfare benefits from pharmaceutical innovation in the U.S. antidepressant market. The antidepressant market has experienced an impressive stream of innovations over the last three decades and available data on antidepressants are exceptionally rich and accurate. While evaluating pharmaceutical innovation in antidepressants, I uncover and address the moral hazard issue that arises due to the existence of prescription drug insurance coverage. The study estimates large patient welfare gains due to innovation and helps explain a detected divergence between social and private patient benefits by the existence of insurance. These findings aid in public policy decision making on health care and pharmaceutical industry concerns.

The study employs an original dataset that consists of annual observations on prices, quantities and drug characteristics for every antidepressant medication sold in the U.S. market from 1980 to 2001. Data also include information on the segmentation of the therapeutic area of antidepressants into different categories of drugs as well as information on branded and generic entry of antidepressants in the U.S. market. Sales data are from the IMS HEALTH Inc. dataset, graciously furnished by Merck & Co.; data on drug characteristics come from combined sources, mainly from the Food and Drug Administration and the Drug Information Handbook; data on patient characteristics were provided by the National Center for Health Statistics. The latter are time-varying demographic data on the distribution of patient income and out-of-pocket prescription drug expenditures.

The first chapter of the dissertation [job market paper] utilizes a structural discrete choice model of hedonic demand to estimate the changes in patient welfare due to antidepressant introduction. To obtain correct substitution patterns the model permits the inclusion of unobserved drug characteristics, which necessitate the use of instrumental variable techniques to correct for price endogeneity. My methodology focuses on the estimation of a full random coefficients multinomial logit model that contributes to the literature in several ways. The model allows for patient observed and unobserved heterogeneity in both willingness-to-pay and taste for branded drugs over generics. Random draws from a joint distribution of income and prescription drug insurance coverage model the observed patient heterogeneity. Draws from an assumed multivariate normal distribution approximate the unobserved heterogeneity of patient preferences.

In addition, the model allows for unobserved patient heterogeneity in the valuation of different drug characteristics that reflects the idiosyncrasy of antidepressant side effects. Estimation uses a simulated method of moments algorithm since demand aggregation involves the computation of multi-dimensional integrals for which there is no analytical solution. The estimated demand parameters are used to calculate own- and cross-price elasticities of demand to describe patient substitution patterns. Consequently, the study makes demand-based assessments of patient willingness-to-pay, with and without insurance, and draws implications for patient welfare. Parameter estimates also allow me to estimate patient willingness-to-pay for hypothetical drugs that might be introduced in the future.

Demand estimates correctly detect marginal disutilities for drug side effects and estimated drug substitution patterns accurately reflect differences in patient tastes for drug attributes. I find a large mean price disutility, which varies with income and insurance demographics. The estimated price sensitivity decreases with patient income and when patients are insured against prescription drug expenditures. Moreover, patients demonstrate a high preference for branded drugs. The wealthier the patients and the more insurance coverage they have, the higher the preference.

Welfare estimation involves the calculation of an upper bound for incremental patient surplus when all the gains obtained are attributed to the innovation in question and a lower bound when the innovative drug is excluded from the choice set at the time of innovation. I obtain large gains for patients, particularly when insurance coverage exists. Relative gains help evaluate the importance of different innovations in the antidepressant market; the innovation of Prozac, which was also the first drug in a new category of antidepressants, offered the highest gains in patient surplus.

The second chapter of the dissertation [work in progress] uses monthly pharmaceutical data from 1996 to 2001 to incorporate patient-level information on prescription drug insurance. This is a major improvement in the model as inclusion of disaggregate data will more accurately address the moral hazard issue. I break down the IMS data by the various distribution channels (for example, non-federal hospitals, private pharmacies and health maintenance organizations) which will aid in explaining the role of institutions in the choice of pharmacological treatment. Survey data that match patient choices directly to patient income and insurance information will then provide even more precise estimates on patient willingness-to-pay and welfare.

In the third chapter, I incorporate IMS data on advertising in the estimation of demand to analyze the effects of marketing efforts by pharmaceutical firms on patient welfare. Preliminary estimation results demonstrate that the market-augmenting demand effects of marketing promotions are more important in the case of depression than the persuasion effects that arise from the promotion of substitution among different antidepressant medications.

Finally, my dissertation also incorporates a more detailed analysis of the workings of the antidepressant market, offers estimation of different functional forms of demand for antidepressants (multinomial logit, different levels of nested logit and a non-random utility model) and compares their performance in the estimated demand parameters as well as the welfare implications they provide. Using similar aggregate data on antibiotics, I adjust the model and estimate demand and welfare in the antibiotic market to make comparisons and evaluate the importance of market specifics on estimation. The comparison will take the methodology a step closer to a model more generally applicable to other drugs and products contributing to methodological innovation in the economics-of-new-goods literature.